Characterization of Rat Liver-specific Methionine Adenosyltransferase Gene Promoter

Methionine adenosyltransferase is a ubiquitous enzyme that catalyzes the only known route of biosynthesis of S-adenosylmethionine, the major methyl group donor in cell metabolism. In mammals, two different methionine adenosyltransferases exist: one is confined to the liver, and the other one is distributed in extrahepatic tissues. In the present study, we report the cloning of the 5′-flanking region of liver-specific methionine adenosyltransferase gene from rat. Two closely spaced sites for transcriptional initiation were identified by primer extension analysis. The major transcription start site was determined to be 29 nucleotides downstream from the putative TATA box. Transient transfection assays of constructs containing sequentially deleted 5′-flanking sequences fused to the luciferase gene showed that rat hepatic methionine adenosyltransferase promoter was able to efficiently drive reporter expression not only in liver-type cells (rat hepatoma H35 cells and human hepatoblastoma HepG2 cells) but also in Chinese hamster ovary cells. Two regions spanning nucleotides −1251 to −958 and −197 to +65 were found to be crucial for the promoter efficiency. The distal upstream region contains elements that positively regulate promoter activity in H35 and HepG2 cells but are ineffective in Chinese hamster ovary cells. Eight protein binding sites were characterized in both regions by DNase I footprinting analysis. Two of these elements, sites A and B, located in the distal region, were found to be essential for the regulation of promoter activity. Electrophoretic mobility shift assays and competition experiments showed that site A is recognized by an NF1 protein. Site B was able to interact with a member of HNF-3 family when nuclear extracts from rat liver and H35 cells were used in the in vitro assay, but an additional binding activity to an NHF1-like protein was obtained with the hepatoma cell extracts. It is suggested that this differential binding can contribute to the cell specificity of promoter function.

S-Adenosylmethionine plays a central role in cellular metabolism, being the major methyl group donor in transmethylation reactions and the source of propylamine moieties for polyamine biosynthesis (1,2). S-Adenosylmethionine is synthesized by transfer of the adenosyl moiety from ATP to the sulfur atom of methionine, in a reaction catalyzed by the enzyme methionine adenosyltransferase (MAT 1 ; ATP:L-methionine S-adenosyltransferase, EC 2.5.1.6) (3). The occurrence of MAT has been extensively studied in a variety of organisms, where different isoenzyme forms have been characterized (2). In mammals, biochemical and molecular cloning studies have revealed the existence of at least two MAT (for a consensus nomenclature for mammalian methionine adenosyltransferase genes and gene products, see Ref. 4). One is selectively expressed in the liver and the other one is distributed in non-hepatic tissues (reviewed in Ref. 5). The presence of a liver-specific isoenzyme is related to the main role of this organ in methionine metabolism. Thus, most of the methionine taken up from the diet is metabolized in the liver, and up to 85% of all transmethylation reactions occur in the liver (5).
Hepatic MAT is a cytosolic homo-oligomeric protein, found as a mixture of tetramers and dimers (6,7). Its expression correlates well with liver growth and differentiation, having been proposed to be a marker of the differentiated state of the hepatocyte (8). There is growing evidence suggesting that this enzyme can be regulated at different levels by a variety of factors and under several pathological conditions. For instance, a serious decrease in the enzyme activity, without a concomitant reduction in the expression of the gene, has been found in several human hepatic disorders (9,10), as well as in different experimental models of liver injury (11,12). In contrast, a marked reduction of MAT gene expression has been reported in human hepatocarcinoma (13) and in a rat model of hypoxiainduced liver injury (14). On the other hand, glucocorticoids (15) and cAMP (8) increase the expression of the gene in rats, whereas insulin blocks the inducing effect of glucocorticoids (8). Altogether, these results suggest that hepatic MAT gene expression is regulated differently under various normal and pathophysiological conditions. The necessity of a strict regulation of the expression of this enzyme has been recently emphasized by the fact that a sustained enhancement in its synthesis is accompanied by a depletion of cellular ATP and NAD, and a greater sensitivity to oxidative cell injury (16).
To study molecular mechanisms underlying the regulation of hepatic MAT expression, we have isolated and characterized the 5Ј-flanking region of the rat MAT gene. The regulatory elements necessary for basal expression have been identified using transient transfections into various cell lines with hepatic MAT promoter-luciferase chimeric genes as well as by in vitro DNase I footprinting analysis. These cis-acting elements are clustered in a promoter proximal region and in a distal region. We finally show that transcriptional activity of hepatic MAT promoter is predominantly dependent on the binding of an NF1-like protein and liver-enriched transcription factors to the promoter distal region.

EXPERIMENTAL PROCEDURES
General Procedures-Standard procedures were used for screening the recombinant genomic library, restriction enzyme mapping, subcloning, DNA labeling, isolation of genomic DNA, Southern transfer, and hybridization to DNA on filters (17,18). Oligonucleotide primers were synthesized on an Applied Biosystems 391 DNA synthesizer using the phosphoramidite method.
Isolation and Characterization of Genomic Clones-A rat genomic DNA library in EMBL-3 SP6/T7 (CLONTECH) was screened using as a probe a 32 P-labeled 580-bp EcoRI/XbaI fragment of the pSSRL cDNA clone (19), which contains 211 bp of the 5Ј-untranslated sequence and the first 369 bp of the rat liver MAT coding region. Five overlapping clones ranging in size between 13 and 18 kb kilobase pairs were isolated from ϳ1 ϫ 10 6 recombinants. Clones were plaque-purified and subjected to Southern analysis. Fragments of interest were subcloned into the plasmid pUC18. DNA was sequenced on both strands by the dideoxy termination method (20), using Sequenase (U. S. Biochemical Corp.) and either pUC/M13 forward and reverse primers or sequence-specific oligonucleotide primers.
Primer Extension Analysis-Poly(A) ϩ RNA used for primer extension studies was prepared from rat liver by oligo(dT)-cellulose chromatography (21). A 24-base oligonucleotide primer complementary to nucleotides Ϫ186/Ϫ163 of the rat liver MAT cDNA (19) was end-labeled with [␥-32 P]ATP using T 4 polynucleotide kinase. Three g of rat liver poly(A) ϩ RNA (or yeast tRNA as negative control) were annealed to 10 6 cpm of the primer and extended with 200 units of Moloney murine leukemia virus reverse transcriptase (Superscript II, Life Technologies, Inc.) as described previously (19). The primer-extended products were analyzed on 7 M urea, 6% polyacrylamide gels, in parallel with sequencing reactions carried out on the genomic subclone using the same primer.
Construction of Promoter-Luciferase Fusion Vectors-Eleven deletion constructs of different length were generated by PCR, using a MAT genomic clone as template. Initially, a DNA fragment spanning 1470 bp of the 5Ј-flanking region was amplified using 3Ј and 5Ј primers corre-sponding to nucleotides ϩ49 to ϩ65 and Ϫ1405 to Ϫ1389, respectively. The purified PCR product was inserted into the SmaI site of pUC18 plasmid. Ten deletions were then constructed by inverse PCR (22), using a vector-specific 5Ј primer and 3Ј primers complementary to nucleotides Ϫ1251 to Ϫ1235, Ϫ1154 to Ϫ1133, Ϫ1134 to Ϫ1118, Ϫ1080 to Ϫ1064, Ϫ958 to Ϫ942, Ϫ727 to Ϫ711, Ϫ527 to Ϫ511, Ϫ375 to Ϫ359, Ϫ193 to Ϫ177, and Ϫ87 to Ϫ71 of the cloned sequence. PCR reactions were performed as described previously (7) but using the thermostable DNA polymerase Dynazyme (Finnzymes Oy, Finland). The inserts of these 11 constructs were digested with KpnI and SalI and subcloned into the corresponding sites of the luciferase promoterless vector pXP1 (23). The identity of the constructs was confirmed by sequencing.
Transient Transfection Analysis-The relative transcriptional activities of MAT promoter fragments were determined by transient transfection analysis in cultured H35, HepG2, and CHO cells. Typically, cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and plated at approximately 3 ϫ 10 5 cells/ 60-mm culture dish 24 h before transfections. The cells were transfected by the calcium phosphate precipitation method (18) with 6 g (CHO) or 15 g (H35, HepG2) of each DNA construct. Five g of the ␤-galactosidase expression vector pCH110 (Pharmacia Biotech) were included as an internal standard of transfection efficiency.
After 18 h, the DNA precipitates were rinsed twice with phosphatebuffered saline, and cells were further grown for 24 h in culture medium. The cells were harvested in reporter lysis buffer (Promega), following the manufacturer's instructions, and the lysate was spun in a microcentrifuge for 15 s. Luciferase and ␤-galactosidase activities were determined as described (17,18). All transfections were conducted in triplicate using at least two different batches of each construct. Activities reported are averaged from three independent experiments.
Preparation of Nuclear Extracts-Nuclear extracts from rat tissues or rat hepatoma H35 cells were prepared as described by Gorski et al. (24) and Andrews and Faller (25), respectively. The final nuclear suspension buffer contained 20 mM HEPES, pH 7.9, 0.4 M KCl, 0.2 mM EDTA, 1.5 mM MgCl 2 , 1 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, 0.5 g/ml leupeptin, 10 g/ml soybean trypsin inhibitor, and 25% glycerol. Nuclear extracts were aliquoted and frozen at Ϫ70°C. Protein concentrations were determined by the Bradford protein assay kit (Bio-Rad).
DNase I Footprinting Analysis-Two 32 P-end-labeled fragments of the 5Ј-flanking region of rat liver MAT gene were generated by PCR amplification of the template GSR4. The primers for these two frag- ments were ϩ49 to ϩ65, Ϫ193 to Ϫ177, Ϫ1012 to Ϫ996, and Ϫ1251 to Ϫ1235. One primer from each pair was 5Ј-end-labeled with [␥-32 P]ATP using T 4 polynucleotide kinase. PCR reactions were performed as described (7). Amplified fragments were purified using Sephacryl S-300 columns (Pharmacia). Approximately 5 ϫ 10 4 cpm of end-labeled DNA fragments were incubated with 10 -50 g of nuclear proteins from rat liver. After 30 min incubation at room temperature, CaCl 2 and MgCl 2 were added to give a final concentration of 0.5 and 1 mM, respectively. DNase I digestions were performed at room temperature for 1 min, using different amounts of enzyme. The reactions were stopped by the addition of 140 l of stop solution containing 2 g of yeast tRNA, 20 mM EDTA, 150 mM sodium acetate, and 1.5 g of proteinase K. Upon phenol extraction and ethanol precipitation, pellets were resuspended in sample dye buffer, and DNA fragments were resolved by electrophoresis in a denaturing 8% acrylamide/urea sequencing gel. The positions of specific DNase I-protected regions were determined by sequencing the fragments according to .

Molecular Cloning of the 5Ј-Flanking Region of Rat Liver
MAT Gene-A rat genomic DNA library in EMBL-3 sp6/T7 was screened with a 580-bp EcoRI/XbaI fragment of the rat liver MAT cDNA clone pSSRL, which comprises 211 bp of the 5Јuntranslated sequence and the first 369 bp of the coding region (19). Five overlapping clones containing inserts ranging in size between 13 and 18 kb were isolated from approximately 1 ϫ 10 6 recombinants. Restriction mapping and Southern analysis showed that all of them contained different lengths of the 5Ј-flanking region. A genomic clone of about 16 kb, designated GRS4, containing approximately 6 kb of the 5Ј-flanking region of the MAT gene was chosen for detailed analysis. To verify that no gross rearrangements had occurred during the cloning process, Southern blots prepared with rat genomic DNA digested with selected restriction endonucleases were hybridized with the same probe used for screening the genomic library. The sizes of the hybridizable fragments were identical to those obtained when Southern analysis of the genomic clone GRS4 was performed under the same conditions (Fig. 1). These re- FIG. 3. Nucleotide sequence and putative regulatory elements of the 5flanking region of rat liver MAT gene. Sequence is numbered relative to the major transcription start site, which is designated as ϩ1. The minor start site is marked with an asterisk. The putative regulatory elements are indicated in bold letters above the underlined sequences.
sults indicate that the cloned DNA segment retains the same sequence organization as in the genomic DNA and suggest that hepatic MAT gene is present as a single copy in the rat genome.
Determination of the Transcriptional Start Sites-To determine the start site of transcription, a primer extension assay was performed using poly(A) ϩ RNA from rat liver, as described under "Experimental Procedures." As shown in Fig. 2, a major product corresponding to a 89-base extended fragment was detected as well as a minor product two nucleotides longer. These products were not detected when the assay was carried out using tRNA. Sequencing reactions performed on the genomic clone GRS4 using the same primer localized the major start site 251 nucleotides from the ATG translation initiation codon and 41 nucleotides upstream of the 5Ј-end of the partial cDNA sequence previously published (19). This base was designated as ϩ1 for numbering the nucleotides in the gene.
Structural Features of the 5Ј-Flanking Region-The nucleotide sequence of rat hepatic MAT promoter region extending to 1557 bp upstream of the transcription initiation site is shown in Fig. 3. The exon sequence corresponding to the first 41 nucleotides of the MAT mRNA, not present in the reported cDNA sequence (19), is also included.
The 5Ј-flanking region shares 88% sequence identity with the 1113-bp promoter region reported for the mouse MAT gene (32). Sequence analysis revealed a putative TATA box located at positions Ϫ29 to Ϫ23, which is in agreement with the preferred position occupied by this element in a typical eukaryotic promoter (33). The canonical CAAT box, usually present around Ϫ80 bp, was not found in this area, but two perfect CAAT motifs were located far away (positions Ϫ379 and Ϫ1514). A number of DNA consensus sequences reported to bind specific trans-acting factors were also present. Thus, two AP-1 binding sites (34,35) were found at positions Ϫ1057 to Ϫ1051 and Ϫ294 to Ϫ288. Two copies of the PEA3 motif (5Ј-AGGAAG-3Ј) (36), an oncogene-, growth factor-, and phorbol ester-responsive element, were located at Ϫ772 to Ϫ767 and Ϫ217 to Ϫ212. The consensus recognition motif for the glucocorticoid response element (TGT(C/T)CT) (37) was represented 4 times at positions Ϫ701 to Ϫ696, Ϫ636 to Ϫ631, Ϫ442 to Ϫ437, and Ϫ66 to Ϫ61. Two interleukin-6 response elements (38), which act as positive elements in several acute phase protein genes, were identified at Ϫ1301 to Ϫ1296, in reverse complement form, and at Ϫ317 to Ϫ312. Two NF-1 binding motifs (5Ј-TGGN 7 CCA-3Ј) (39) were found at Ϫ1218 to Ϫ1206 and Ϫ630 to Ϫ618, the latter overlapping with a glucocorticoid response element. Finally, the analysis also revealed several DNA elements that matched the consensus binding sequences for liver-enriched transcription factors, such as HNF-1 (27) (at Ϫ1144 to Ϫ1138), HNF-3 (40) (at Ϫ1128 to 1122 and Ϫ581 to Ϫ575), and HNF-4 (41) (at Ϫ159 to Ϫ154 and Ϫ79 to Ϫ73).
Functional Analysis of the 5Ј-Flanking Region of Rat Liver MAT Gene-To delineate the sequences that drive MAT expression, a series of deletion mutants extending from Ϫ1405 to ϩ65 bp was generated by PCR and cloned into the promoterless luciferase expression vector pXP1 (Fig. 4A). A similar plasmid containing the RSV promoter was used as a positive control for transfection. These chimeric constructs were transiently transfected into two cell types of liver origin, the rat hepatoma H35 cell line and the human hepatoblastoma HepG2 cell line, as well as into CHO cells. In all transfection experiments, a vector expressing ␤-galactosidase was included as internal control. Transient expression of luciferase activity showed that the MAT-luciferase vectors were expressed in the three cell types. The pattern of transcriptional activity of the chimeric constructs was very similar in H35 and HepG2 cells (Fig. 4B). Only background expression was observed following transfection with the pXP1 luciferase vector without any promoter fragment. The highest luciferase activity, almost 14 and 20% of RSV-luciferase activity in H35 and HepG2, respectively, was observed with the Ϫ1405/ϩ65 construct. Removal of the fragment Ϫ1405 to Ϫ1251 had little effect on the promoter efficiency. However, successive 5Ј deletions from Ϫ1251 to Ϫ87 bp resulted in a marked decrease in luciferase activity, indicating the presence of a positive-acting region between Ϫ1251 and Ϫ958, that increases transcriptional activity about 3-fold. Deletions of sequences from Ϫ958 to Ϫ727, Ϫ727 to Ϫ527, Ϫ527 to Ϫ375, and Ϫ375 to Ϫ193 did not significantly affect luciferase activity, suggesting that no relevant additional elements in-volved in basal transcription are contained in this area. Further deletion from Ϫ193 to Ϫ87 promoted a 2-fold decrease, suggesting the presence of positive regulatory element(s) in this area. The resulting construct (Ϫ87/ϩ65), which contains the minimal promoter elements including the TATA box, produced 10 and 18% maximal activity in H35 and HepG2 cells, respectively.
The profile of luciferase expression in CHO cells was quite different from that obtained in the liver-type cells (Fig. 4C). The constructs Ϫ1405/ϩ65, Ϫ1251/ϩ65, and Ϫ958/ϩ65 yielded a similar luciferase activity. Therefore, as judged by the results obtained in the hepatic cells, the region comprised between Ϫ1251 and Ϫ958 might contain tissue-specific regulatory elements that do not account for the promoter activity in CHO cells. On the other hand, and in contrast with the pattern observed in H35 and HepG2 cells, sequences from Ϫ958 to Ϫ727 and Ϫ375 to Ϫ195 seem to contain elements that positively regulate transcription in CHO cells.
DNase I Footprinting Analysis of MAT 5Ј-Flanking Region-From the results of deletion analysis, it can be concluded that the essential elements governing MAT expression in hepatic cells lie between nucleotides Ϫ1251 to Ϫ958 and Ϫ193 to ϩ65. To identify these potential cis-acting elements, DNase I footprinting analysis was performed using both DNA fragments and nuclear proteins isolated from rat liver. As shown in Fig.  5A, five DNase I-protected areas, designated I to V, were generated on both strands within the promoter proximal region (Ϫ193 to ϩ65) at increasing amounts of protein. The nucleotide sequences of these footprints are depicted in Fig. 6. Footprint I is a large protected area that covers putative TATA box elements. The second protected region extends from Ϫ46 to Ϫ70 and includes a significant homology to an AP1 recognition site. Footprint III shows no sequence similarity to any of the known protein-binding sites, as revealed by a search in transcription factor site data bases (42,43). The region between Ϫ150 and Ϫ96 (footprints IV and V) adjoins sequences bearing some resemblance to several putative transcription factor binding sites, including cAMP response element, AP1, and AP4. Interestingly, this region also contains two tandem copies of the 5Ј-GTCGAA-3Ј motif, which does not match any consensus recognition site for known transcription factors.
Three prominent DNase I-protected areas and several hypersensitive sites were identified upon incubation of a DNA fragment extending the promoter distal region from Ϫ1251 to Ϫ996 with liver nuclear extracts (Fig. 5B). Footprint A spans a putative NF1 binding site, and footprint B (Ϫ1132 to Ϫ1155) includes both HNF1 and HNF3 consensus sites homologies. Footprint C presents some sequence similarity to NF1 and MYOD recognition sites. Several DNase I-hypersensitive bands were also detected in this promoter region, even at low protein input (indicated with asterisks in Fig. 5B). These could also reflect the binding of nuclear proteins, which would make this area more accessible to the action of the DNase I.
To approach the identification of elements that could represent binding sites for liver-enriched transcription factors, DNase I footprinting assay was also performed using nuclear extracts from other rat tissues. As shown in Fig. 7A, incubation of the promoter proximal region with nuclear proteins from liver, kidney, or lung yielded similar DNase I footprints, showing a minimal variation in DNase I-hypersensitive cleavage accompanying footprint IV. It is therefore suggested that ubiquitous trans-acting factors are interacting with this promoter area. The promoter distal region exhibited a distinct DNase I protection pattern upon incubation with the nuclear extracts from different sources. Thus, while footprint A was detectable with all nuclear extracts, elements B and C were clearly protected only by nuclear proteins from liver, although the former seemed to be weakly protected by kidney extract (Fig. 7B). These findings suggest that element A represents a binding site for a ubiquitous transcription factor, whereas elements B and C interact with liver-enriched factors.
Relative Contribution of Protein-binding Sites in the Distal Promoter Region to Transcription Activity-Since the region spanning nucleotides Ϫ1251 to Ϫ958 appeared to be crucial for the promoter activity in hepatic cells, we next focused on the analysis of the functional significance of the protein binding sites detected by footprinting in this area. For this purpose, 5Ј deletion mutants extending up to each footprint were carried out; the constructs were transfected in H35 and CHO cells, and the transcription capacity was assessed (Fig. 8). Removal of the fragment Ϫ1251 to Ϫ1154 containing the area corresponding to footprint C promoted a small decrease (about 20%) of luciferase activity. However, an additional deletion of sequences corresponding to element B (Ϫ1154 to Ϫ1134) resulted in about a 2-fold increase in reporter gene transcription, indicating the presence of elements that negatively influence the promoter function. Finally, upon deleting a fragment comprising also the footprint A, the promoter efficiency was reduced 3-fold, rendering a relative luciferase activity similar to that exhibited by the deletion construct Ϫ958/ϩ65 (Fig. 4). On the other hand, and in agreement with the data shown in Fig. 4, no differences in luciferase activity were detected upon transfection of these deletion constructs in CHO cells.
Characterization of Nuclear Protein Binding to Sequences within Footprints A and B-Since sequences residing within footprints A and B appeared to be of major functional relevance, the identity of the potential factors that interact with these elements was explored by gel mobility shift analysis. Double-stranded synthetic oligonucleotides (named FPA and FPB; see "Experimental Procedures") encompassing each of the two DNase I footprinted sites were used as probes and incubated with nuclear extracts from rat liver and H35 cells. To assess the specificity of each protein binding site, a series of double-stranded oligonucleotides was used as competitors in the binding reactions.
As shown in Fig. 9, the radiolabeled FPA probe formed two closely spaced DNA-protein complexes with rat liver and H35 nuclear extracts. The specificity of DNA binding was established by the ability of a 50-fold molar excess of the same unlabeled probe to compete for complex formation. An NF1 consensus oligonucleotide efficiently competed, even better than the homologous FPA, for binding with both nuclear extracts, whereas no competition was observed with oligonucleotides containing HNF-1, HNF-3, C/EBP, or AP1 consensus sequences. This result suggests that the protein forming these complexes is a member of the NF1 family or an NF1-like protein. Footprint B displays sequence homology to consensus binding sites for the liver-enriched factors HNF-1 and HNF-3. Accordingly, competition experiments were carried out using unlabeled oligonucleotides corresponding to binding sequences for these factors, as well as for C/EBP. Two major gel shift complexes were detected when the FPB probe was incubated with nuclear extracts from rat liver; the upper one fully competed by a 50-fold molar excess of the unlabeled oligonucleotide (Fig.  10). Addition of a mutant form of FPB sequence (MtFPB) or a C/EBP consensus element failed to compete with protein binding. However, an HNF-3 oligonucleotide markedly reduced the complex formation. This was also competed, although to a lesser extent, by an HNF-1 consensus binding site. Nuclear extracts from H35 cells formed a different pattern of shifted complexes when incubated with radiolabeled FPB. Thus, one complex of similar electrophoretic mobility to the major com-plex formed with rat liver extracts was detected, as well as an additional complex of slower mobility. Both shifted bands could be abolished by addition of excess unlabeled probe but not by the mutated FPB sequence or C/EBP consensus element. The formation of the upper complex was abolished by an excess of HNF-1 oligonucleotide, which also showed a weak cross-competition with the lower complex. Conversely, addition of an HNF-3 consensus sequence did not affect the formation of the upper complex but efficiently competed with the lower complex. In fact, the latter exhibited a similar behavior in competition experiments with both cell extracts. In summary, these results suggest that in both cell extracts MAT promoter FPB element is interacting with an HNF3-like protein, whereas in the H35 extracts, the same element also binds an HNF1-like factor. DISCUSSION Among mammalian tissues, the liver is a major site of production of MAT, needed to metabolize most of the methionine taken from the diet and to provide key metabolites for the cell. The crucial role of the liver-specific MAT (1,5,44,45) requires a precise regulation of its expression under basal conditions and in response to different stimuli. In this study, we report the cloning and initial characterization of the 5Ј-flanking region of the rat hepatic MAT gene, providing insights into the mechanisms that regulate its expression.
The gene for the rat liver-specific MAT is present as a single copy in the genome, as has also been reported for its human (10) and mouse (32) counterparts. The major start site of transcription was determined to be 29 nucleotides downstream of the putative TATA sequence, at the same site as that for the mouse MAT gene (32), further supporting the assignation of this nucleotide as the cap site. However, unlike the mouse gene, this site corresponds to a T residue, a nucleotide not usually found at the transcription start site in the eukaryotic promoters analyzed (33).
Functional analysis by transient transfections has shown that the promoter for the rat hepatic MAT is active not only in liver-type cells but also in a non-hepatic cell such as CHO. This finding raises the possibility that the liver-restricted expression of the endogenous gene is not mediated by the action of tissue-specific transcription factors. In such cases, the lack of its expression in the extrahepatic tissues could be due to situations known to stably repress the transcription of many tissue-specific genes, such as DNA methylation or genome organization (i.e. compaction of the chromatin) (46,47). By analyzing a series of 5Ј-end deletions, two major regulatory regions have been identified, a promoter proximal region (Ϫ193 bp relative to the transcription start site) and a far upstream area extending nucleotides Ϫ1251 to Ϫ958. Most notably, promoter elements located in the latter produced the highest reporter activity upon transfection in H35 or HepG2 cells.
It is interesting to note that major differences can be found when the functional analysis of the 5Ј-flanking region of rat liver MAT gene is compared with that reported for its mouse counterpart, despite the high sequence conservation in the 1113-bp overlapping regions. Basically, in the mouse promoter, the 5Ј cis-acting elements that produce the highest transcriptional activity are located between nucleotides Ϫ365 to Ϫ145, and a negatively acting region has been identified between Ϫ518 to Ϫ366 (32). Nevertheless, it is not known whether the crucial distal elements found in the rat promoter are also present in its mouse counterpart, since in the latter this region has not been analyzed.
The promoter proximal region of rat hepatic MAT appears to contain positive elements commonly active in both hepatic-type and CHO cells. Thus, the region spanning positions Ϫ87 to ϩ65 exhibited similar relative transcriptional activity in the three cell lines tested, probably owing to the action of basic regulatory sequences such as TATA box. In fact, three protected elements, one of them covering the putative TATA box, were detected in this DNA fragment by using the in vitro footprinting assay, and they appear to represent binding sites for ubiquitous trans-acting factors, as judged by analysis with nuclear extracts from different sources. Two additional footprinted areas were found between nucleotides Ϫ193 and Ϫ87. These elements also seem to be of functional relevance since its deletion produced a significant decrease in promoter activity. The identity of the proteins binding to this proximal region is the subject of further investigation.
The distal region is crucial for basal and tissue-specific promoter activity. Thus, its removal led to a drastic reduction in the reporter activity upon transfection in H35 and HepG2 cells but had no effect in CHO cells. The structure-function relationship of this area was studied by a combination of DNase I footprinting analysis, transient transfections, and gel shift assays. With this approach, three relevant cis-acting elements (A-C) were defined. Sequences spanning sites A and C bind factors that positively regulate transcription of MAT promoter, with a predominant role for site A. Gel retardation and competition experiments have provided evidence that this element interacts with a member of the NF1 family, a finding that is consistent with the initial identification by sequence homologies. Our results suggest that NF1, which usually acts as a positive transcription factor (30,48), plays a key role in directing MAT expression.
On the other hand, site B appears to bind a factor that negatively modulates transcriptional activity in H35 cells. Surprisingly, an oligonucleotide containing this element showed, in gel shift assays, different specific interactions with nuclear extracts from rat liver and hepatoma H35 cells. Thus, it was able to recognize a protein present in both cell extracts that, as judged by competition experiments, appears to be a member of the liver-enriched transcription factor HNF-3 family. However, an additional binding activity, corresponding to a potential HNF-1 transcription factor, was detected with the H35 extract. It is interesting to note that site B contains the sequence 5Ј-CGTATTGATTAAC-3Ј, which is 90 and 84% identical to the consensus HNF-3 (49) and HNF-1 (27) binding elements, respectively. Therefore, both factors could be partially interacting with the same core sequence. This assumption is supported by the fact that a mutated form of element B disrupting this sequence was unable to compete with the wild-type probe for the formation of both protein complexes. The absence of HNF-1 binding activity in the liver extract is, however, difficult to explain, since hepatic tissue expresses high levels of this transcription factor (50,51). One possible explanation derives from the fact that de-differentiated hepatoma cells do not express HNF-1 but present a protein called variant HNF-1 (vHNF-1 or HNF1␤, Refs. 52 and 53). Although both factors can interact with the same target elements, it has been suggested that the sequence of the binding site may modulate the affinities of HNF-1 proteins, thus providing another level of regulation (54). Therefore, vHNF-1 or a related protein present in hepatoma cells might somehow display a higher affinity for element FIG. 9. Gel mobility shift assay of nuclear proteins interacting with the FPA sequence proteins binding to the MAT promoter FPA sequence. Labeled oligonucleotide probe corresponding to DNase I-footprinted region A was incubated with 5 g of nuclear extracts from rat liver or rat hepatoma H35 cells. For specific competition, a 50-fold molar excess of unlabeled oligonucleotide corresponding to binding sites for HNF-1, HNF-3, c/EBP, NF1, and AP1 were used (see "Experimental Procedures" for the sequences of the oligonucleotides). The binding reactions were separated by electrophoresis in a native 6% polyacrylamide gel. DNA-protein complexes are indicated by arrowheads.
FIG . 10. Differential binding of nuclear proteins from rat liver and H35 cells to the FPB sequence. Gel shift analysis of the interactions of FPA with rat liver and H35 nuclear proteins. Competition was carried out with a 50-fold molar excess of the indicated unlabeled oligonucleotides. An FPB mutant (MtFPB; see "Experimental Procedures") was included as competitor. Arrowheads indicate the specific binding complexes. B than HNF-1 and then compete with HNF-3 binding. The finding that element B negatively regulates MAT transcription is in agreement with this hypothesis, since HNF1 is a positive regulatory factor (50,51), and vHNF-1 has been shown to be involved in the negative regulation of some liver-specific genes (55). In this context, it should be mentioned that steady-state levels of liver-specific MAT mRNA are lower in rat hepatoma H35 cells than in hepatocytes (15). A similar pattern of reduced expression has been also detected for the human hepatic mRNA in the hepatoma HepG2 cell line, 2 a finding that does not completely agree with a previous report suggesting a complete lack of transcription of the liver MAT gene in these cells (13).
It is well documented that the successful expression of several liver-specific genes requires a combined action of ubiquitous factors, such as NF-1, and liver-enriched factors (50,51,56,57). HNF-3 proteins function as transcriptional activators (58), and their cooperative interaction with NF-1 has been reported to be needed, among others, for the hepatocyte-specific expression of albumin gene (59) or for full activity of the promoter of C4b binding protein (60). A combined action of both factors might also account for the proper expression of rat liver MAT. In fact, their respective sequence elements are tandemly arrayed and separated by only 16 nucleotides. Therefore, the finding that in hepatoma cells an additional protein can interact with the HNF-3 element of MAT promoter raises the possibility that such binding could interfere with the transactivation ability of this transcription factor. In this line, the results presented here could explain the different basal expression of MAT in rat liver and hepatoma cells and provide more insights into the mechanisms governing the expression of liver-specific genes in differentiation.